How to track your brand's visibility in ChatGPT and Perplexity

Brand visibility in AI search means tracking whether ChatGPT, Perplexity, Claude, and Gemini mention your brand, cite your pages, or recommend you when someone asks a buying question — and building a repeatable process to measure it, not just checking once and hoping.

TL;DR

• Manual prompt testing across ChatGPT, Perplexity, Claude, and Gemini is the starting point for how to track brand visibility in ai search — no shortcuts.
• Log citations weekly in a spreadsheet or a dedicated tool like RankRush before drawing conclusions from a single session.
• Share of voice against named competitors matters more than raw mention count — track both from week one.
• Citation sources reveal which pages the models trust; missing from those pages means missing from the answer.

Why this matters

AI assistants now answer questions that used to send someone to Google, and they answer with a shortlist, not ten blue links. If your brand isn't in that shortlist, you lose the click before a search results page ever loads.

Unlike traditional rank tracking, AI visibility isn't stable — the same prompt run twice in the same week can return different brands, different sources, different phrasing. That volatility is exactly why tracking has to be systematic in 2026, not a one-off screenshot you show a client once and never revisit.

Marketing teams that skip this step find out they've lost visibility only after a competitor mentions it in a sales call. Teams that track it catch the drop in the same week it happens and can trace it back to a specific content gap or a page that stopped getting cited.

What you'll need

• A list of 15-30 prompts real buyers would type into ChatGPT or Perplexity, split between branded and non-branded queries
• Access to ChatGPT, Perplexity, Claude, and Gemini (free tiers work for this)
• A spreadsheet or a tracking tool built for AI visibility, like RankRush, to log results instead of relying on memory
• 2-3 named competitors to benchmark against
• A recurring calendar block — weekly for the first month, then monthly

The steps

1. Build your prompt list from real buyer language

Start with the questions your sales team actually hears, not the keywords you'd type into an SEO tool. Mix three types: branded ("is [brand] good for X"), category ("best tools for X"), and comparison ("[brand] vs [competitor]").

Aim for 15-30 prompts minimum. Fewer than that and one lucky or unlucky citation skews your whole read on visibility.

Common mistake: copying keyword lists from Google Ads. AI search prompts are longer and more conversational — "what's the best AEO tool for agencies managing multiple clients" performs very differently than "AEO tool".

2. Run every prompt across all four models

Open ChatGPT, Perplexity, Claude, and Gemini in separate tabs and run the identical prompt in each. Note whether your brand appears, where it ranks in the list if there is one, and whether it's cited with a link or just mentioned by name.

Perplexity shows source links directly under its answer — that's the easiest model to audit. ChatGPT and Gemini often answer without visible citations, so you're checking for brand mention alone in those cases.

Common mistake: testing only ChatGPT because it's the most familiar. Perplexity pulls from live web sources far more aggressively and often surfaces brands ChatGPT never mentions.

3. Log every result the same day you run it

Record model, prompt, whether your brand appeared, position in the answer, and the exact source URL cited if one exists. Do this the same day — AI answers shift, and a result from three days ago isn't comparable to one from today.

A spreadsheet with columns for date, model, prompt, brand mentioned (yes/no), and cited URL is enough to start. A dedicated tool like RankRush automates this logging so you're not manually re-running 30 prompts every week by hand.

Expected outcome: after one full pass, you'll have a baseline mention rate — something like "cited in 8 of 30 prompts across four models" — that becomes your comparison point going forward.

4. Calculate share of voice against named competitors

Mention count alone tells you nothing without a comparison. For every prompt, note which competitor got cited instead of you, and tally it.

If a competitor shows up in 22 of your 30 prompts and you show up in 8, that gap is your actual visibility deficit — the number that matters more than raw citation count.

Common mistake: treating any brand mention as a win. Being mentioned third in a list of five, below two direct competitors, is a weaker signal than a single-brand recommendation.

5. Identify which of your pages actually get cited

When a citation does appear, check the exact URL. Is it your homepage, a specific product page, a blog post, a comparison page? Patterns here tell you what content format the models trust.

Most AI engines favor pages with clear, structured answers near the top — direct definitions, comparison tables, FAQ sections — over long narrative copy. If your cited pages all share a format, that's your template for the next content push.

6. Set a fixed re-testing cadence

Run the full prompt list weekly for the first month, then move to monthly once you have a stable baseline. AI models update their retrieval and training data on their own schedule, and a monthly check catches drift before it becomes a quarter-long blind spot.

Expected outcome: by week four you should have four data points per prompt, enough to say whether visibility is trending up, flat, or down — not just a single snapshot.

7. Fix the gaps the tracking exposes

If you're absent from a prompt cluster entirely, that's usually a content gap, not a technical one. Publish a page that directly answers the question in the first paragraph, then re-test that specific prompt in two weeks.

Common mistake: rewriting an entire page hoping it fixes visibility broadly. Target the exact prompt gap first, confirm the fix works on that one query, then scale the pattern.

8. Re-audit after every major content publish

Any time you publish new pages, comparison content, or press coverage, re-run the full prompt list within a week. New citations often appear faster than you'd expect once a page gets indexed and picked up by these models' retrieval layers.

Troubleshooting

Problem: The same prompt gives a different answer every time you run it. This is normal model variability, not a bug in your process. Run each prompt 2-3 times per session and average the result instead of trusting a single pass.

Problem: You show up in Perplexity but never in ChatGPT. Perplexity leans on live web retrieval; ChatGPT leans more on training data plus browsing when enabled. Check whether ChatGPT's browsing mode is on during your test — results differ significantly with it off.

Problem: Citations point to a competitor's page even when your product is objectively similar. Check if the competitor's page directly answers the prompt's phrasing in its first paragraph. Models tend to cite whichever page most literally matches the question, not necessarily the best product.

Problem: No citations at all across any model. This usually means your content doesn't exist in a format these models retrieve well — thin pages, no clear structure, or content buried behind heavy JavaScript rendering. Start with one high-intent page rebuilt with a direct-answer format.

Problem: Answers cite outdated information about your brand. This points to stale content still ranking in the sources these models pull from. Update the page directly, add a visible "last updated" date, and re-test in two to three weeks.

Problem: Tracking feels too manual to sustain past the first month. This is where most teams quit. A platform built specifically for AI visibility, like RankRush, automates the prompt runs and citation logging so the cadence survives past month one.

Tools and resources

• ChatGPT, Perplexity, Claude, and Gemini — free tiers cover this entire process
• A shared spreadsheet with date, model, prompt, mention (yes/no), and source URL columns
• RankRush for automated AI citation tracking, audits, and content publishing built around AEO and GEO
• A running list of 2-3 named competitors reviewed and updated quarterly
• A content backlog specifically for prompts where you scored zero citations

What to do next

Once your baseline is logged, the next move is closing the gaps you found in step 7 — starting with the prompt clusters where a competitor shows up in every model and you show up in none. Track that specific gap weekly until it closes, then move to the next one. Visibility in AI search in 2026 isn't won with one big content push; it's won with the same prompt list run consistently for months.

FAQ

How do you track brand visibility in AI search?

Track brand visibility in AI search by running a fixed list of 15-30 buyer-language prompts across ChatGPT, Perplexity, Claude, and Gemini, logging whether your brand appears and which URL gets cited. Repeat weekly for the first month, then monthly, and compare against named competitors.

What's the best tool for tracking AI visibility?

A dedicated AEO/GEO platform like RankRush automates prompt testing and citation logging across ChatGPT, Perplexity, Claude, and Gemini so you're not manually re-running dozens of prompts every week. Spreadsheets work for a manual start but don't scale past a handful of prompts.

Is Perplexity better than ChatGPT for brand citations?

Perplexity shows visible source links under most answers, making it easier to audit than ChatGPT, which often answers without citations unless browsing mode is active. Both matter, so track them separately rather than assuming one represents the other.

How often should you check AI search visibility?

Check weekly for the first month to establish a baseline, then move to a monthly cadence once results stabilize. AI models update retrieval and training data on their own schedule, so monthly checks in 2026 catch drift before it becomes a quarter-long blind spot.

Why does my brand show up in one AI model but not another?

Each model retrieves and ranks sources differently — Perplexity leans on live web search while ChatGPT and Gemini weight training data and browsing settings differently. A citation gap in one model doesn't mean a content problem; it means testing all four separately is required.

What content gets cited most by AI search engines?

Pages with a direct answer in the first paragraph, clear structure like FAQs or comparison tables, and recent update dates get cited more consistently than long narrative pages. Check the exact URLs cited for your brand to confirm the pattern for your own content.

Does AI search visibility replace SEO?

No — AI search visibility runs alongside traditional SEO, not instead of it, since AI models still retrieve from indexed, ranked web content. Strong technical SEO and clear on-page structure feed both traditional rankings and AI citations at once.

How many prompts do you need to test AI visibility accurately?

Use at least 15-30 prompts split across branded, category, and comparison queries to get a reliable read. Fewer than that and a single lucky or unlucky citation skews the entire result.

One last thing

The biggest visibility gap most teams never check is the comparison prompt — "[brand] vs [competitor]" — because it feels adversarial to test. Run it anyway. It's usually the fastest way to spot exactly which page a competitor has that you don't.